This code extracts the facial landmark coordinates obtained from mediapipe into a NumPy array of pixel coordinates, then creates a gif with the 3D coordinates of it. This could be a great place to start if you're wanting to create a 3D filter of your own or to create an animation of yourself or others.
This code is adapted from a longer tutorial on facial landmark detection that you can find here. It's also important to note that mediapipe only works with Python 3.7-3.10 at the moment, so you will need to be running an appropriate version of Python for this to work.
